Improvements to IEnvironmentActivationService to use hidden terminals

[DonJayamanne]

Today we use IEnvironmentActivationService to generate environment variables for activated environments. E.g. when running python code on behalf of user as a background process.

We do this today by activating the environment and capturing the environment variables.
For this we get the activation commands (using a default shell) and run it as a process.

A better approach is:

• Create a hidden terminal
• Get activation command and send to terminal
• Dump variables to text file (instead of stdout).

Solves

• unicode issues (today scraping stdout unicode)
• Activates in terminal (thats what some activation command expect, i.e. to be run within a shell).

Discussed with @karthiknadig (helps DAP) @int19h /cc

FYI - I have started some work on this.

[int19h]

I think this could even skip the terminal altogether, and just spawn the activation script directly (all it needs is to ensure that it runs under the same shell that the actual terminal will be).

[DonJayamanne]

I think this could even skip the terminal altogether, and just spawn the activation script directly (all it needs is to ensure that it runs under the same shell that the actual terminal will be).

Agreed. But using the terminal is easier as we don't need to worry about finding paths to shells.
The activation scripts will run in the exact same environment that would other wise run when users run their code. We also have the added advantage of displaying the shell for diagnostic purposes if something goes wrong.

Also this ensures startup scripts are correctly loaded in the shells (something conda requires and initializes when using conda init).

On a side note, a tty interface is required by either pipenv or pyenv. Can't remember which one it was.

[rchiodo]

I believe this is working. At least it does for me. @IanMatthewHuff didn't you say you had a problem with activation with a new environment?

[IanMatthewHuff]

@rchiodo no. My issue was different, I'm pretty sure, my issue was with our saving of the last successful jupyter environment I think.

[rchiodo]

Thanks. Validated then.